Speech Recognition Pinyin Correction for Polyphonic Ambiguity

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current offline speech recognition systems face challenges in accurately converting voice inputs to text and understanding user intentions due to issues like polyphonic characters and pronunciation ambiguities, leading to suboptimal recognition results.

Innovation Solution

A speech recognition method that involves converting initial recognition results into candidate character pinyin strings, determining sentence pinyin strings, and performing pinyin correction using a pinyin correction database to improve accuracy, thereby enhancing the understanding of user intentions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If offline speech recognition is used to convert voice to text locally, then processing speed and privacy are improved, but recognition accuracy deteriorates due to polyphonic characters and pronunciation ambiguities

Engineering Contradiction:
Improverecognition accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the speech recognition process into multiple stages: initial recognition, character-level pinyin generation, sentence-level pinyin assembly, and correction. This segmentation allows each stage to handle specific aspects of accuracy improvement independently, resolving the contradiction between accuracy and complexity by organizing the complex task into manageable segments.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary actions by generating candidate character pinyin strings for each character before assembling them into sentence-level pinyin strings. This preliminary character-level processing enables more accurate matching against the correction database, improving overall recognition accuracy while maintaining systematic complexity management.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If polyphonic characters are handled using traditional recognition methods, then processing simplicity is maintained, but recognition accuracy deteriorates due to pronunciation ambiguities

Engineering Contradiction:
Improverecognition accuracyVSAvoidprocessing simplicity
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The patent changes the representation parameter from direct character recognition to pinyin string representation. By converting characters to their pinyin equivalents and generating multiple candidate pinyin strings for polyphonic characters, the system transforms the recognition problem into a pinyin matching problem, significantly improving accuracy for polyphonic characters while maintaining operational simplicity through automated processing.

Inventive Principle:
Principle #35Parameter changes

3Measurement precision

If a generic word database is used for correction, then database complexity is reduced, but correction accuracy deteriorates due to inability to handle polyphonic characters properly

Engineering Contradiction:
Improvecorrection accuracyVSAvoiddatabase size
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent segments the correction process at the character level, generating candidate pinyin strings for each character separately before assembling them into sentence-level corrections. This segmentation allows the correction database to be searched more effectively, improving correction accuracy without requiring an excessively large database by breaking down the correction task into manageable character-level operations.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP4027337B1Speech recognition method and apparatus, electronic device and storage medium
Publication Date: 2024.02.14 APOLLO INTELLIGENT CONNECTIVITY (BEIJING) TECH CO LTD
  • EP4027337B1 patent drawingFigure 1
  • EP4027337B1 patent drawingFigure 2
  • EP4027337B1 patent drawingFigure 3

AI summary

The present disclosure provides a speech recognition method, a speech recognition apparatus, an electronic device and a storage medium, and relates to the field of computer technology, in particular to the field of artificial intelligence technology such as speech recognition and natural language processing. The method includes: obtaining an initial recognition result by performing a speech recognition on a sentence to be recognized; obtaining at least one candidate character pinyin string corresponding to each character in the initial recognition result; determining at least one sentence pinyin string corresponding to the initial recognition result based on the at least one candidate character pinyin string corresponding to the character; and generating a pinyin-corrected recognition result by performing pinyin correction on the initial recognition result based on the at least one sentence pinyin string. Therefore, the accuracy of the speech recognition result is improved.